Huong Yong Ting ¹, Yong Wen Daniel Tan ¹, Boung Yew Simon Lau ¹

Affiliations
1 School of Computing, University College of Technology Sarawak, 96000 Sibu, Sarawak, MY

Abstract

Objectives: Badminton performance analysis and profiling are essential steps in badminton coaching in order to identify the strengths and weaknesses of an athlete. Methods: In this paper, we investigated and identified the key potential and limitations of the Microsoft Kinect sensor for badminton performance analysis, particularly on novice badminton player. A survey was conducted during badminton event of SUKMA in order to determine which strokes are important to novice level player, where participants include coaches from different states of Malaysia. The essential strokes were then analyzed by Microsoft Kinect sensor. Findings: The survey results indicated that there are four main strokes to be mastered by novice level player, such as clear, net, lift and smash. Moreover, the key selected badminton strokes as identified by expert coaches such as forehand crosscourt lift, backhand touch net, backhand lift, forehand lift, forehand push net, backhand clear, backhand push net, forehand touch net and forehand clear can be measured and analyzed accurately and consistently with Microsoft Kinect sensor. However, the sensor measurements are limited for badminton strokes such as static smash, jump smash and overhead forehand clear. The major reason for such limitation is mainly due to occlusions and loss of acquisition data due to fast moving motion. Improvement: Therefore, the current performance analysis algorithm using skeleton information will be incorporated with color information in future to resolve the occlusion issue.

Keywords

Badminton, Depth Map, Kinect, Performance Analysis, Performance Profiling.

Full Text

   

Huong Yong Ting ¹, Yong Wen Daniel Tan ¹

Affiliations
1 School of Computing, University College of Technology Sarawak, 96000 Sibu, Sarawak, MY

Abstract

Objectives: In this paper, we extended our previous novel lossless compact view invariant compression technique, namely Range of Movement Index (RoMI) by fusing an adaptive module. Methods: An adaptive module is proposed to fuse with the RoMI to elevate the technique to become left-right handed invariant. The module will firstly identify the label of the normalized RoMI value from a particular range in order to determine left or right side. Subsequently, the adaptive mapping functions are utilized to perform left to right or right to left mappings using the identified label. Findings: Generally, badminton players can be categorized into three different handed: mainly right-handed players and left handed-players and rarely the ambidextrous players. In our previous technique, the RoMI can only benchmark or perform computerized badminton movement quality comparison based on the handedness of a badminton player. In specific, the benchmarking mechanism is left-right handed variant, i.e., left-handed player with left-handed player and right-handed player with right-handed player. This limitation will increase the effort to benchmark badminton players’ movement quality with different handedness of badminton player. The proposed adaptive module enables the comparison of computerized stroke movements between different players with different handedness. As such, this new method will identify the labels of the normalized RoMI and performs adaptive mapping to match with the reference handedness to produce a more consistent benchmarking of different handedness badminton players. Improvement: The ability to benchmark different handedness badminton players enables the system to be adopted by a larger range of badminton players and further simplify data collection and analysis procedures.

Keywords

Adaptive, Badminton Stroke, Kinect, Movement Comparison, Range of Movement Index (RoMI)

Full Text